Release v1.23.0

Added

• Bounded jobs: JobsOptions.MaxActiveStepsPerJob (or Job.MaxActiveSteps) keeps a sliding window of active steps and starts the next one as each step reaches an outcome. A resumed job rebuilds its window from storage. Off by default.
• Step attempts: JobsOptions.MaxStepAttempts, StepRetryBackoff and MaxStepRetryBackoff, with Job.ShouldRetryStep deciding per failure. A failed attempt is scheduled again by the job with exponential backoff and counts as a failure only when its attempts run out. Default is one attempt.
• IJobsManager.RetryFailedSteps(jobId) and Job.CanRetryFailedSteps run the failed steps of a job that completed with failures again. Opt-in per job type; resuming such a job is unchanged.
• Optional step leases: JobsOptions.StepLeaseTimeout makes running steps record a heartbeat. The watchdog restarts a step only when its owner silo has left the cluster and its lease has expired; steps from older releases are left alone for one lease period.
• IJobStepStorage.CountByStatus(jobId) as a default interface member, implemented for MongoDB, SQL and in-memory storage. Persisted steps gain an optional Attempt and Lease; existing data reads as attempt 1 with no lease.
• JobsOptions.JobWatchdogInterval, StepDeliveryMaxAttempts, StepDeliveryRetryDelay and MaxStepDeliveryRetryDelay.

Changed

• IJobsManager.Stop now returns Result<StopJobError> instead of Task; update callers that await it as a plain task if they need the outcome.
• A running job is watched: it counts its steps in storage and finalizes itself when every step is done, counting Failed and CompletedWithFailure as failed, instead of staying Running when a step's report never arrived.
• A step persists its terminal status first and then delivers it to the job, retrying with backoff until acknowledged. A delivery that never succeeds no longer marks the step failed; the job accepts each outcome once per step and attempt.

Fixed

• Job.Stop no longer reports a stop that was not persisted. It restores the job's state and answers the new StopJobError.CouldNotPersistStop.